SPC: What is Statistical Process Control?

Introduction

Statistical Process Control (SPC) is a method designed to monitor and control product or process variability using statistical tools.

The main idea is that it is less expensive to check the quality of products or processes by periodically examining a small sample rather than inspecting every single product. Additionally, in situations where the testing method destroys the product, conducting a 100% inspection is simply not feasible.

SPC allows to supervise the production process based on predetermined rules. We take a small sample (e.g., 3 pieces of product) from the process, then evaluate it and apply the result to a so-called control chart as a new measurement. If this new result indicates that everything is "normal," we continue production. However, if the result indicates that some abnormal disturbance has occurred, then we regulate the process. That is the idea of process control.

SPC may be applied to both process parameters (inputs) and product characteristics (process outputs). If technically feasible, place more emphasis on monitoring process parameters to achieve more repeatable, high-quality products.

SPC primarily uses control charts, along with supporting tools like: histograms, statistical tests, run charts, data stratification, process capability indices etc.

This article is an introduction to statistical process control (SPC), so be aware that certain aspects have been simplified.

So.. let's dive in!

History of SPC

• 1920+. The origins of SPC can be traced back to the pioneering work of Dr. Walter A. Shewhart at Bell Telephone Laboratories. Shewhart developed the concept of the control chart, a core tool in SPC, to monitor and control manufacturing processes.
• 1931. Shewhart wrote a book[1] in which he introduced the concept of using statistical methods to differentiate between normal process variations and special causes that require investigation. The book "Economic Control of Quality of Manufactured Product" placed the foundation for modern quality control practices.
• 1931+. Shewhart's ideas gained wider recognition. His colleague, Dr. W. Edwards Deming, further promoted his work and strongly advocated using SPC for quality improvement[2].
• 1940+. SPC methods were introduced in the U.S. during World War II, particularly in the production of munitions and other critical supplies where quality was essential.
• 1950+. After World War II, W. E. Deming traveled to Japan. He introduced Statistical Process Control to Japanese industries[3], significantly influencing Japan's post-war economic recovery. Japanese manufacturers implemented these principles into their production processes, significantly improving quality and efficiency. At that time, Mr. Kaoru Ishikawa popularised[4] basic quality tools:
  • Control chart
  • Histogram
  • Check sheet
  • Data stratification
  • Cause-and-effect diagram known as Ishikawa diagram
  • Pareto chart
  • Scatter diagram
• 1980+. From the 1980s onward, industries in the USA and Europe began to recognize SPC as a vital tool for quality improvement. W. E. Deming had a significant role in this movement through his book "Out of the Crisis"[5] and by providing training to the American industry. His work helped raise awareness in the West about the benefits of SPC. Additionally, the U.S. government contributed to this shift by introducing the Malcolm Baldrige National Quality Award in 1987[6], which encouraged businesses to adopt various quality improvement techniques, including SPC. West Europe took a similar approach by introducing The European Foundation for Quality Management (EFQM) in 1989[7].
• 1990+. Motorola introduced the Six Sigma methodology, incorporating SPC techniques to minimize variation and improve processes. The "Big Three" U.S. automakers, General Motors, Ford, and Chrysler, developed the "SPC Reference Manual" as part of the QS-9000 quality management system. It was a very important step in standardizing these practices across the automotive industry.
• 2000+. SPC methods have gained popularity across various industries. The AIAG SPC Manual and VDA Volume 4 are among the most commonly used guides in the automotive sector. At the same time, ISO standards such as ISO 11462-1, ISO 7870 (series), and ISO 22514 were released in parallel, providing additional global guidelines for implementing statistical process control.

Variation in the process

In the context of Statistical Process Control (SPC), variation refers to the differences or fluctuations that occur in a manufacturing process, resulting in inconsistencies in the final product.

Variation is inherent in all manufacturing processes. It can result from fluctuations in the quality of materials, gradual machine wear and tear, fluctuations in environmental conditions, errors in equipment settings, and disturbances and vibrations that occur during production.

From the SPC point of view, variability is the result of common causes and sometimes special causes.

Common causes

Common causes, also known as "natural causes" or "random causes," are inherent in the process and are almost constantly present. The changes are usually small, and their average value is nearly constant over time. So, we can say with some simplification that common causes are small, random factors constantly disrupting the process - but in a predictable way.

Examples of common causes in a hole drilling process in a metal are as follows:

• Tool wear (normal)
• Machine vibration (inherent due to machine design)
• Material properties variation (within spec)
• Coolant flow variability
• Environmental conditions (minor, typical changes)

Reducing the impact of these factors is possible by making fundamental changes in the process, such as changing the machine to a more accurate one, introducing air conditioning on the production floor, implement automation, etc.

Special causes

Special causes are undesirable events that cause significant deviations from normal process variability. They can usually be attributed to specific causes that need to be reduced or eliminated.

Examples of special causes in a hole drilling process in a metal are as follows:

• Tool damage
• Machine excessive vibration (upcoming failure, misalignment)
• Material properties out of spec
• The coolant flow nozzle or filters are partially clogged.
• Sudden environmental conditions change.

SPC charts role

The purpose of SPC charts is to distinguish normal causes (natural, inherent fluctuations) from special causes (excessive changes, failures, problems) in the monitored process. Thus, it can be said that the control chart helps to control the process, indicating when it is actually necessary to regulate the process and when it is not necessary to react.

The following are key elements of a control chart that serve the role:

• Process average line. The average line, also known as the 'central line,' represents the expected process performance. It helps detect shifts or trends in the process, indicating when the process is performing differently than expected.
• Control limit lines. These lines, typically set at ±3 standard deviations from the process average, provide a quick indication when the process goes out of control limits. They help identify specific causes that are significantly affecting the process.

Some special causes can occur within the control limits and go unnoticed. I recommend using the Nelson rules to identify these "hidden" special causes. These rules help spot patterns like trends or shifts, even if they are still within the control limits. This ensures that no special causes go unnoticed.

SPC charts types

Statisticians developed many statistical process control (SPC) chart types over time. The most common are:

• X-R. Average (X) and range (r) chart. The most common chart in the manufacturing industry.
• X-s. Average (X) and standard deviation (s) chart.
• X_m-R. Median (Xm) and range (R).
• I-MR. Individual (I) and moving range (MR).
• p. Percent or Fraction Defective Parts Chart
• np. Number Defective Parts Chart.
• c. Count of Nonconformities Chart.
• u. Nonconformities per Unit Chart.

There are also other less common control charts such as:

• CUSUM. Cumulative Sum Chart.
• EWMA. Exponentially Weighted Moving Average Chart.

Capability indexes

Capability indices estimate a process's limits relative to its specification limits. The most common are:

• Cp, Cpk. Process Capability
• Pp, Ppk. Process Performance
• Cm, Cmk. Machine Capability

Additionally, Measurement System Analysis (MSA) uses Gauge Capability (Cg, Cgk) in so-called "Type I study".

Manufacturers use capability indexes to understand and communicate how well a process, machine, or gauge performs relative to specification limits. The manufacturing industry has established minimum limits for these indicators; for example, a Cp or Cpk value greater than 1.67 is now a common requirement.

SPC standards

Several ISO and industry specyfic standards describe SPC charts and capability calculations. Following is a list of popular standards:

• AIAG SPC. "Statistical Process Control (SPC)" Reference Manual
• VDA Volume 4. "Quality Assurance in the Process Landscape"
• AESQ AS13006. "Process Control Methods"
• ISO 22514 (series). "Statistical Methods in Process Management - Capability and Performance"
• ISO 7870 (series). "Control Charts"
• ISO 8258 "Shewhart Control Charts"

Summary

To use SPC charts effectively, you first need to understand their true purpose, learn how to build them, and then know how to use them properly.

Implementing SPC can be challenging. Training is just the beginning; applying SPC charts in real-time is often much harder. It can be difficult in some organizations to encourage operators to stop a process to make adjustments before a product problem arises, especially if the focus is mainly on production quantity.

It's also crucial to understand a characteristic being monitored. Is it normally distributed? If not, remember that some SPC calculations, like Cp/Cpk indices, assume a normal distribution and may cause misunderstanding.

SPC is a powerful statistical method in an engineer's toolbox, and every Super Engineer 😊 should understand its importance in professional work.